Experience from an operational map-reduce cluster reveals that outliers significantly prolong job completion. Mantri culls outliers using cause- and resource-aware techniques. Its strategies include smart restart of outliers, network-aware placement of tasks and protecting outputs of valuable tasks. Deployment in Bing’s production cluster and extensive trace-driven simulation indicate that Mantri is 3.1x more effective than the existing state-of-the-art in improving job completion times.